Getting started
Using gpyrn
should be simple if you are familiar with Python. Just import the
package directly or each of the three sub-packages
The covfunc
package provides covariance functions (kernels) to be used for the
GPRN nodes and weights. meanfunc
provides the mean functions to use for a
given dataset. Note that, in the GPRN model, the nodes and weights are
independent GPs with mean zero; these mean functions will apply to the output
datasets. The heavy-lifting is done by the mean-field approximation that is
implemented in meanfield
.
As described in the examples, the typical use will be to
instantiate a meanfield.inference
object passing in the observed datasets, and
then defining the GPRN components (nodes, weights, and means). So typically you
would do something like
# load data...
# create an inference object
gprn = meanfield.inference(N_NODES, time_array, *outputs_and_errors)
# define GPRN components
nodes = [
covfunc. ...
]
weights = [
covfunc. ...
]
means = [
meanfunc. ...
]
jitters = [...]
gprn.set_components(nodes, weights, means, jitters)
after which you can calculate gprn.ELBO
or optimize the parameters with
gprn.optimize()
.